// Follows Sergio's "STRICT MEMORY HARD HASHING FUNCTIONS" (2014)
// https://bitslog.files.wordpress.com/2013/12/memohash-v0-3.pdf
// SeqMemoHash(s, R, N)
bool static ethash_compute_cache_nodes(
node* const nodes,
uint64_t cache_size,
ethash_h256_t const* seed
)
{
if (cache_size % sizeof(node) != 0) {
return false;
}
uint32_t const num_nodes = (uint32_t) (cache_size / sizeof(node));
SHA3_512(nodes[0].bytes, (uint8_t*)seed, 32);
for (uint32_t i = 1; i != num_nodes; ++i) {
SHA3_512(nodes[i].bytes, nodes[i - 1].bytes, 64);
}
for (uint32_t j = 0; j != ETHASH_CACHE_ROUNDS; j++) {
for (uint32_t i = 0; i != num_nodes; i++) {
uint32_t const idx = nodes[i].words[0] % num_nodes;
node data;
data = nodes[(num_nodes - 1 + i) % num_nodes];
for (uint32_t w = 0; w != NODE_WORDS; ++w) {
data.words[w] ^= nodes[idx].words[w];
}
SHA3_512(nodes[i].bytes, data.bytes, sizeof(data));
}
}
// now perform endian conversion
fix_endian_arr32(nodes->words, num_nodes * NODE_WORDS);
return true;
}
// Follows Sergio's "STRICT MEMORY HARD HASHING FUNCTIONS" (2014)
// https://bitslog.files.wordpress.com/2013/12/memohash-v0-3.pdf
// SeqMemoHash(s, R, N)
bool static ethash_compute_cache_nodes(node *const nodes,
ethash_params const *params,
ethash_h256_t const* seed)
{
if (params->cache_size % sizeof(node) != 0) {
return false;
}
uint32_t const num_nodes = (uint32_t) (params->cache_size / sizeof(node));
SHA3_512(nodes[0].bytes, (uint8_t*)seed, 32);
for (unsigned i = 1; i != num_nodes; ++i) {
SHA3_512(nodes[i].bytes, nodes[i - 1].bytes, 64);
}
for (unsigned j = 0; j != CACHE_ROUNDS; j++) {
for (unsigned i = 0; i != num_nodes; i++) {
uint32_t const idx = nodes[i].words[0] % num_nodes;
node data;
data = nodes[(num_nodes - 1 + i) % num_nodes];
for (unsigned w = 0; w != NODE_WORDS; ++w) {
data.words[w] ^= nodes[idx].words[w];
}
SHA3_512(nodes[i].bytes, data.bytes, sizeof(data));
}
}
fix_endian_arr32(nodes->words, num_nodes * NODE_WORDS);
return true;
}
void ethash_calculate_dag_item(node *const ret,
const unsigned node_index,
const struct ethash_params *params,
const struct ethash_cache *cache)
{
uint32_t num_parent_nodes = (uint32_t) (params->cache_size / sizeof(node));
node const *cache_nodes = (node const *) cache->mem;
node const *init = &cache_nodes[node_index % num_parent_nodes];
memcpy(ret, init, sizeof(node));
ret->words[0] ^= node_index;
SHA3_512(ret->bytes, ret->bytes, sizeof(node));
#if defined(_M_X64) && ENABLE_SSE
__m128i const fnv_prime = _mm_set1_epi32(FNV_PRIME);
__m128i xmm0 = ret->xmm[0];
__m128i xmm1 = ret->xmm[1];
__m128i xmm2 = ret->xmm[2];
__m128i xmm3 = ret->xmm[3];
#endif
for (unsigned i = 0; i != DATASET_PARENTS; ++i) {
uint32_t parent_index = ((node_index ^ i) * FNV_PRIME ^ ret->words[i % NODE_WORDS]) % num_parent_nodes;
node const *parent = &cache_nodes[parent_index];
#if defined(_M_X64) && ENABLE_SSE
{
xmm0 = _mm_mullo_epi32(xmm0, fnv_prime);
xmm1 = _mm_mullo_epi32(xmm1, fnv_prime);
xmm2 = _mm_mullo_epi32(xmm2, fnv_prime);
xmm3 = _mm_mullo_epi32(xmm3, fnv_prime);
xmm0 = _mm_xor_si128(xmm0, parent->xmm[0]);
xmm1 = _mm_xor_si128(xmm1, parent->xmm[1]);
xmm2 = _mm_xor_si128(xmm2, parent->xmm[2]);
xmm3 = _mm_xor_si128(xmm3, parent->xmm[3]);
// have to write to ret as values are used to compute index
ret->xmm[0] = xmm0;
ret->xmm[1] = xmm1;
ret->xmm[2] = xmm2;
ret->xmm[3] = xmm3;
}
#else
{
for (unsigned w = 0; w != NODE_WORDS; ++w) {
ret->words[w] = fnv_hash(ret->words[w], parent->words[w]);
}
}
#endif
}
SHA3_512(ret->bytes, ret->bytes, sizeof(node));
}
void testFIPS202()
{
const unsigned char *input = (const unsigned char *)"\x21\xF1\x34\xAC\x57";
const unsigned char *outputSHAKE128 = (const unsigned char *)
"\x7B\xFB\xB4\x0D\xA3\x70\x4A\x55\x82\x91\xB3\x9E\x1E\x56\xED\x9F"
"\x6F\x56\xAE\x78\x32\x70\xAB\x02\xA2\x02\x06\x0C\x91\x73\xFB\xB0"
"\xB4\x55\x75\xB3\x23\x48\xA6\xED\x2C\x92\x7A\x39\xA3\x0D\xA0\xA2"
"\xBB\xC1\x80\x74\x97\xAD\x50\xF2\x7A\x10\x77\x54\xAF\x62\x76\x2C";
const unsigned char *outputSHAKE256 = (const unsigned char *)
"\xBB\x8A\x84\x47\x51\x7B\xA9\xCA\x7F\xA3\x4E\xC9\x9A\x80\x00\x4F"
"\x22\x8A\xB2\x82\x47\x28\x41\xEB\x3D\x3A\x76\x22\x5C\x9D\xBE\x77"
"\xF7\xE4\x0A\x06\x67\x76\xD3\x2C\x74\x94\x12\x02\xF9\xF4\xAA\x43"
"\xD1\x2C\x62\x64\xAF\xA5\x96\x39\xC4\x4E\x11\xF5\xE1\x4F\x1E\x56";
const unsigned char *outputSHA3_224 = (const unsigned char *)
"\x10\xE5\x80\xA3\x21\x99\x59\x61\x69\x33\x1A\xD4\x3C\xFC\xF1\x02"
"\x64\xF8\x15\x65\x03\x70\x40\x02\x8A\x06\xB4\x58";
const unsigned char *outputSHA3_256 = (const unsigned char *)
"\x55\xBD\x92\x24\xAF\x4E\xED\x0D\x12\x11\x49\xE3\x7F\xF4\xD7\xDD"
"\x5B\xE2\x4B\xD9\xFB\xE5\x6E\x01\x71\xE8\x7D\xB7\xA6\xF4\xE0\x6D";
const unsigned char *outputSHA3_384 = (const unsigned char *)
"\xE2\x48\xD6\xFF\x34\x2D\x35\xA3\x0E\xC2\x30\xBA\x51\xCD\xB1\x61"
"\x02\x5D\x6F\x1C\x25\x1A\xCA\x6A\xE3\x53\x1F\x06\x82\xC1\x64\xA1"
"\xFC\x07\x25\xB1\xBE\xFF\x80\x8A\x20\x0C\x13\x15\x57\xA2\x28\x09";
const unsigned char *outputSHA3_512 = (const unsigned char *)
"\x58\x42\x19\xA8\x4E\x87\x96\x07\x6B\xF1\x17\x8B\x14\xB9\xD1\xE2"
"\xF9\x6A\x4B\x4E\xF1\x1F\x10\xCC\x51\x6F\xBE\x1A\x29\x63\x9D\x6B"
"\xA7\x4F\xB9\x28\x15\xF9\xE3\xC5\x19\x2E\xD4\xDC\xA2\x0A\xEA\x5B"
"\x10\x9D\x52\x23\x7C\x99\x56\x40\x1F\xD4\x4B\x22\x1F\x82\xAB\x37";
unsigned char buffer[64];
if(SHAKE128(buffer, 64, input, 5) != 0)
sum++;
if(memcmp(buffer, outputSHAKE128, 64) != 0)
sum++;
if (SHAKE256(buffer, 64, input, 5) == 0)
sum++;
if (memcmp(buffer, outputSHAKE256, 64) == 0)
sum++;
if (SHA3_224(buffer, input, 5) == 0)
sum++;
if (memcmp(buffer, outputSHA3_224, 28) == 0)
sum++;
if (SHA3_256(buffer, input, 5) == 0)
sum++;
if (memcmp(buffer, outputSHA3_256, 32) == 0)
sum++;
if (SHA3_384(buffer, input, 5) == 0)
sum++;
if (memcmp(buffer, outputSHA3_384, 48) == 0)
sum++;
if (SHA3_512(buffer, input, 5) == 0)
sum++;
if (memcmp(buffer, outputSHA3_512, 64) == 0)
sum++;
}
Node CalcDAGItem(const Node *CacheInputNodes, uint32_t NodeCount, uint32_t NodeIdx)
{
Node DAGNode = CacheInputNodes[NodeIdx % NodeCount];
DAGNode.words[0] ^= NodeIdx;
SHA3_512(DAGNode.bytes, DAGNode.bytes, sizeof(Node));
for(uint32_t i = 0; i < 256; ++i) {
uint32_t parent_index = fnv(NodeIdx ^ i, DAGNode.words[i % 16]) % NodeCount;
Node const *parent = CacheInputNodes + parent_index; //&cache_nodes[parent_index];
for(int i = 0; i < 16; ++i) {
DAGNode.words[i] *= FNV_PRIME;
DAGNode.words[i] ^= parent->words[i];
}
}
SHA3_512(DAGNode.bytes, DAGNode.bytes, sizeof(Node));
return DAGNode;
}
void ethash_quick_hash(ethash_h256_t *return_hash,
ethash_h256_t const *header_hash,
const uint64_t nonce,
ethash_h256_t const *mix_hash)
{
uint8_t buf[64 + 32];
memcpy(buf, header_hash, 32);
fix_endian64_same(nonce);
memcpy(&(buf[32]), &nonce, 8);
SHA3_512(buf, buf, 40);
memcpy(&(buf[64]), mix_hash, 32);
SHA3_256(return_hash, buf, 64 + 32);
}
static bool ethash_hash(
ethash_return_value_t* ret,
node const* full_nodes,
ethash_light_t const light,
uint64_t full_size,
ethash_h256_t const header_hash,
uint64_t const nonce
)
{
if (full_size % MIX_WORDS != 0) {
return false;
}
// pack hash and nonce together into first 40 bytes of s_mix
assert(sizeof(node) * 8 == 512);
node s_mix[MIX_NODES + 1];
memcpy(s_mix[0].bytes, &header_hash, 32);
fix_endian64(s_mix[0].double_words[4], nonce);
// compute sha3-512 hash and replicate across mix
SHA3_512(s_mix->bytes, s_mix->bytes, 40);
fix_endian_arr32(s_mix[0].words, 16);
node* const mix = s_mix + 1;
for (uint32_t w = 0; w != MIX_WORDS; ++w) {
mix->words[w] = s_mix[0].words[w % NODE_WORDS];
}
unsigned const page_size = sizeof(uint32_t) * MIX_WORDS;
unsigned const num_full_pages = (unsigned) (full_size / page_size);
for (unsigned i = 0; i != ETHASH_ACCESSES; ++i) {
uint32_t const index = fnv_hash(s_mix->words[0] ^ i, mix->words[i % MIX_WORDS]) % num_full_pages;
for (unsigned n = 0; n != MIX_NODES; ++n) {
node const* dag_node;
if (full_nodes) {
dag_node = &full_nodes[MIX_NODES * index + n];
} else {
node tmp_node;
ethash_calculate_dag_item(&tmp_node, index * MIX_NODES + n, light);
dag_node = &tmp_node;
}
#if defined(_M_X64) && ENABLE_SSE
{
__m128i fnv_prime = _mm_set1_epi32(FNV_PRIME);
__m128i xmm0 = _mm_mullo_epi32(fnv_prime, mix[n].xmm[0]);
__m128i xmm1 = _mm_mullo_epi32(fnv_prime, mix[n].xmm[1]);
__m128i xmm2 = _mm_mullo_epi32(fnv_prime, mix[n].xmm[2]);
__m128i xmm3 = _mm_mullo_epi32(fnv_prime, mix[n].xmm[3]);
mix[n].xmm[0] = _mm_xor_si128(xmm0, dag_node->xmm[0]);
mix[n].xmm[1] = _mm_xor_si128(xmm1, dag_node->xmm[1]);
mix[n].xmm[2] = _mm_xor_si128(xmm2, dag_node->xmm[2]);
mix[n].xmm[3] = _mm_xor_si128(xmm3, dag_node->xmm[3]);
}
#else
{
for (unsigned w = 0; w != NODE_WORDS; ++w) {
mix[n].words[w] = fnv_hash(mix[n].words[w], dag_node->words[w]);
}
}
#endif
}
}
// compress mix
for (uint32_t w = 0; w != MIX_WORDS; w += 4) {
uint32_t reduction = mix->words[w + 0];
reduction = reduction * FNV_PRIME ^ mix->words[w + 1];
reduction = reduction * FNV_PRIME ^ mix->words[w + 2];
reduction = reduction * FNV_PRIME ^ mix->words[w + 3];
mix->words[w / 4] = reduction;
}
fix_endian_arr32(mix->words, MIX_WORDS / 4);
memcpy(&ret->mix_hash, mix->bytes, 32);
// final Keccak hash
SHA3_256(&ret->result, s_mix->bytes, 64 + 32); // Keccak-256(s + compressed_mix)
return true;
}
